Document Classification Using Click Information to Filter Spam
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing techniques for detecting spam web pages are ineffective due to their design to circumvent automatic detection methods, leading to a high percentage of irrelevant results in search engine queries, particularly for popular topics like entertainment, which hampers user experience and search engine efficiency.
Innovation Solution
A system that classifies documents based on user inputs and click information to identify and filter out off-topic or spam pages by using feature vectors and Bloom filters to adjust relevancy scores and rankings, ensuring that only relevant pages are prioritized in search results.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If automatic detection techniques are used to identify spam pages, then detection speed is improved, but detection accuracy deteriorates because spam pages are designed to circumvent automatic techniques
Solution Approach 1:
The patent introduces click information as an intermediary data source to bridge the gap between automatic detection and spam page evasion techniques. By using user click behavior data as a mediator, the system can identify spam pages without relying solely on automatic analysis techniques that spam pages are designed to circumvent. The click information serves as an independent verification layer that reveals the true nature of pages regardless of their optimization for automatic detectors.
Solution Approach 2:
The system implements feedback loops where click information from user interactions is continuously collected and used to refine spam detection. The detection process uses click data to identify patterns of spam behavior, then applies these insights to improve future detection accuracy. This feedback mechanism allows the system to adapt to evolving spam techniques while maintaining high detection accuracy without sacrificing speed.
2Productivity
If spam pages are allowed in search results, then search engine crawling efficiency is improved, but user experience deteriorates due to high percentage of irrelevant results
Solution Approach 1:
The system performs preliminary classification of pages as spam or non-spam using click information before pages are included in search results. By detecting and flagging spam pages in advance through analysis of click behavior patterns, the system can exclude these pages from search results without affecting the crawling process. This preliminary action ensures that crawling efficiency is maintained while user experience is protected from spam content.
Solution Approach 2:
The patent segments the web page population into distinct categories (spam and non-spam) based on click information analysis. This segmentation allows the search engine to treat different types of pages differently in search result generation. By dividing the page set and applying different handling rules to spam versus legitimate pages, the system maintains crawling efficiency for all pages while ensuring only quality pages appear in user-facing search results.
3Ease of operation
If aggressive spam filtering is applied, then user experience is improved, but loss of information increases as legitimate pages may be incorrectly classified
Solution Approach 1:
The system applies partial filtering by using click information to identify pages with suspiciously high click patterns that suggest spam behavior. Rather than filtering all pages with high click counts, the system applies nuanced analysis to distinguish between legitimately popular pages and spam pages designed to attract clicks. This partial action approach filters out true spam while preserving legitimate high-traffic pages, improving user experience without causing excessive loss of information.
Data Source
AI summary
One embodiment of the present invention provides a system that automatically classifies documents (such as web pages) based on user inputs. During operation, the system obtains a “classified” set of documents which are classified as relating to a specific topic. The system also obtains queries related to the specific topic. These queries produce “query results” which enable the user to access documents related to the query. The queries also include “click information” which specifies how one or more users have accessed the query results. The system uses this click information to identify documents in the classified set of documents which are not related to the specific topic or are off-topic. If such documents are identified, the system shifts the identified documents so that they are regarded as off-topic and/or spam, and removes the identified documents from the classified set of documents.